Prof G Markets
00:32:18 · Josh Brown
[00:32:18] S4: The per unit cost of compute is going down. It's actually crashing. The thing is the usage is going up by so much that the overall dollar amount being spent on compute does not fall, it goes up. So the per token cost is plunging like it should. This is the history of technology. That's good. The thing is the services are so popular and in so much demand that while the per token cost falls, the sheer explosion in everyday usage is rapidly outpacing that so that the dollar cost of compute is not falling. It's rising, but you'd make the argument that's actually a good thing because it shows how sustainable this bull market in AI demand should be. What do you think about that?
00:33:17 · Ed Elson, Josh Brown
[00:33:17] S3: I think that the fact that the token spend... The the spend on the tokens on the on the the, application layer end is going down is a problem for Anthropic and OpenAI. So I think compute costs per token per token compute costs coming down is gonna be very important for them. And I think the more that they can manage figuring out how to how to get as much compute as possible and paying the least amount for it, that's what they have to do. I think they're in a difficult position to do that. But if if if the the per token compute costs are coming down for them, that's a great thing. The only thing that's the problem is if the startups who are buying the the the tokens or buying the LLMs usage from them, if that per token spend is also going down, which it is, then that is also gonna be a problem because it means that their revenues are gonna start to slow. And that's what I worry about, and I worry that that considering the extent to which their their business model is reliant on startups that are primarily funded with VC dollars. In fact, based on ramps data, they show that 80%, of the spending on OpenAI and Anthropic is coming from 1% of the companies. And within that 1%, 80% of them are venture backed. It seems as though they're reliant on capital continuing to rush in in Silicon Valley, in the VC industry. [00:34:39] S4: They're reliant on the stock market to stay buoyant.
00:34:39 · Josh Brown, Ed Elson
[00:34:39] S4: They're reliant on the stock market to stay buoyant. [00:34:41] S3: I I guess here... Here's where I land. I think this is where probably you and I agree, Josh. And you point this point this out about Masa's son, which is like, I think you said he's a riverboat gambler. Like, he's making these very bold bets. And I think that's what this is. And I think as a as a result, all I can say is it could be an incredibly valuable company, but I think they could also get caught in a very bad situation. And there is significant risk that they don't keep up, with their costs, and eventually they could go bankrupt. [00:35:14] S4: I know you guys have spoken to Steve Eisman. We had him on our show recently. Maybe he said the same thing. But I do think that this is the the crux of the matter. The Fortune 500 CEOs and CFOs two years ago said, do AI to their employees. What do you mean? I don't know. I'm on I'm on a stage every week somewhere, and everyone's talking about these amazing things that they're doing with AI. Go do some AI. Okay. [00:35:44] S3: Do it.
00:35:45 · Josh Brown
[00:35:45] S4: And in the early days of this, it's crazy to be talking about the early days. I'm talking about like eighteen months ago. The way that enterprises were being built was like, you know, SaaS, like, oh, here's how much per head per month. Go at it. And then the token the the token use just started getting out of control, like guys pressing enter and just having the machine running for forty eight hours over the weekend while they were in Cape Cod. Like, that's obviously not gonna... The the the guy from Uber came out in March and said we blew through the $20.26 token budget, already for the year in in the first quarter. So now you have two things happening at once, and this is the real bear case. Number one, there's gonna be some sort of budget on AI use company wide, and they're gonna have to make decisions like, okay, this department gets X dollars in budget for AI, that department only gets Y dollars because what the hell are we doing here? We're not running experiments, we're running a business. Okay. So I wouldn't say rationing token use, I would say rationalizing token use. Okay, so that's one thing. The second thing is the weight, the open weight models, the free models, and just this idea that there actually are no switching costs and a fortune 500 business with a sophisticated IT department and an army of consultants, they look at the quarterly numbers, they say, why are we, you know, 7¢ per share behind expectations? Where do we trace it back to, oh shit, look how much money we just paid Anthropic? Isn't there a way that we can swap in some of the, like we can use a harness and the harness will be a governor on cost and it'll decide, okay, for this leg of the trip, we need a Mercedes Benz, but that's only 10%. For the rest of the 90% of the trip, we'd be fine with an ox cart. And all of a sudden, you start seeing companies get smarter about the efficiency with which they utilize these tools. That's obviously bearish if you're trying to sell the premium furthest out on the frontier model at a premium price. So to me, that's a really reasonable bear case. And probably you don't have to go so far as say it's a price war. Maybe it is. But like it's the opening innings of people being a little bit more governed about how they're spending money and where the money goes.
00:38:23 · Ed Elson
[00:38:23] S3: We'll be right back after the break. And if you're enjoying the show so far, send it to a friend, and please follow us on YouTube or on Spotify or wherever you get your podcasts.
00:38:40 · Scott Galloway
[00:38:40] S0: Support for the show comes from VCX, the public ticker for private tech. For generations, American companies have moved the world forward through their ingenuity and determination. And for generations, everyday Americans be a part of that journey through perhaps the greatest innovation of all, The US stock market. It didn't matter whether you were a factory worker in Detroit or a farmer in Omaha, anyone could own a piece of the great American companies. But now that's changed. Today, our most innovative companies are staying private rather than going public. The result is that everyday Americans are excluded from investing and getting left further behind while a select few reap all the benefits until now. Introducing VCX, the public ticker for private tech, now available wherever you buy stocks. VCX by Fundrise gives everyone the opportunity to invest in the next generation of innovation, including the companies leading the AI revolution, space exploration, defense tech, and more. Visit getvcx.com for more info. That's getvcx.com. Carefully consider the investment material before investing, including objectives, risks, charges, and expenses. This and other information can be found in the fund's prospectus at getvcx.com. This is a paid sponsorship. Support for the show comes from Perplexity Computer. If you run an ecommerce business, this might seem familiar. You start your morning by opening, say, five to 10 tabs. Shopify, Google Ads, Klaviyo, maybe a few half baked spreadsheets you've been meaning to organize better. Then you do the dance of making all those tabs work together. Before you know it, it's already lunchtime and you've spent your day pulling information together from a half a dozen different places. Perplexity computer can help simplify things. Instead of opening every dashboard yourself, computer connects to the tools you already use and built a command center that gives you a complete picture of your business. It pulls together your revenue, ad spend, margin, inventory, email performance, and more, all before you've even had your coffee. The best part is that it doesn't tell you what's happening. It actually gets to work, and instead of being locked into one AI model, it can send your project to the right model for the job. For a limited time, only our listeners can try Perplexity Computer for free. Go to pplx.ai/propgmarkets.